The Anthropic prospectus has not been filed publicly yet, but its contents are out. Reuters and the Financial Times have both reviewed the draft Anthropic prospectus, and on 28 and 29 September 2026 they reported what it says: a $42 billion net loss for 2025, revenue up twelvefold, $518 billion of future computing commitments, and a warning that the company’s own AI could pose “catastrophic or existential risks to humanity”.
TechCrunch summed up the oddity well. It is “a strange position for any company to be in — warning that its product could end humanity, while making some of its earliest investors and employees extraordinarily wealthy in the process.” Backers reportedly hope the listing will value Anthropic above $2 trillion, which would make it one of the largest IPOs ever.
This article walks through the numbers in the Anthropic prospectus, separates the accounting charge from the operating loss, checks the growth and valuation arithmetic, sets out what the risk factors actually say, and explains what it all means for businesses that depend on Claude.
Table of contents
- What the Anthropic Prospectus Shows at a Glance
- How the Anthropic Prospectus Became Public
- The Losses in the Anthropic Prospectus
- Growth: The Other Half of the Anthropic Prospectus
- Compute Costs and $518 Billion of Commitments in the Anthropic Prospectus
- Two Customers and No Long-Term Contracts
- The Warning That Its AI Could End Humanity
- Why Risk Factors Fill 80 Pages of the Anthropic Prospectus
- Pace the Frontier Meets the Public Markets
- What the Anthropic Prospectus Implies About a $2 Trillion Valuation
- What the Anthropic Prospectus Means for Businesses Using Claude
- What Happens Next to the Anthropic Prospectus
- Anthropic Prospectus FAQs
- References
What the Anthropic Prospectus Shows at a Glance
The table collects the headline figures reported so far, with the outlet that reported each one. Every figure comes from the draft Anthropic prospectus as described by Reuters, the FT or Fortune; none has yet been published by Anthropic.
| Item | Figure | Reported by |
|---|---|---|
| 2025 revenue | Nearly $4.6bn, up twelvefold (1,088%) | Reuters, Fortune |
| 2025 operating loss | $8.06bn, from $2.98bn in 2024 | Reuters, Fortune |
| 2025 net loss | About $42bn, including a ~$34bn accounting charge | Reuters |
| Total operating expenses | $12.65bn | Reuters |
| Compute and infrastructure | $7.33bn, three times 2024 | Reuters |
| Future commitments | $518bn for cloud, compute and infrastructure | Reuters |
| Cash and short-term investments | $20.28bn at 31 December 2025 | Reuters, Fortune |
| Q2 2026 revenue | $11.5bn; second straight quarter of adjusted operating profit | FT |
| Customer concentration | Two customers made up nearly a quarter of 2025 revenue | FT, Reuters |
| Risk factors | About 80 of 261 pages, against 48 on the business | Reuters |
How the Anthropic Prospectus Became Public
Strictly, it has not. Anthropic submitted a confidential draft registration statement to the SEC on 1 June 2026, as Reuters reported earlier this month, and a confidential draft is not published until the company files publicly. The Decoder reports that the company “sent the S-1 filing to a small group of partners over the past few days”. Reuters was first with the financial details on Monday 28 September, and the FT followed with the risk factors. Fortune describes the document as “leaked”.
What we checked
A search of the SEC’s EDGAR full-text database on 29 September returned no Anthropic registration statement, only fund holdings reports that list Anthropic shares. So every figure in this article is the draft Anthropic prospectus as described by journalists who have seen it. Some numbers may change before the public filing, and the public version will contain full financial statements that the reports only sample.
The Losses in the Anthropic Prospectus
The headline $42 billion net loss in the Anthropic prospectus is real, but it is not what it sounds like. According to Reuters, about $34 billion of it is an accounting charge “that reflected an increase in the estimated value of financing that could eventually turn into Anthropic shares, rather than money the company spent running its business.”
Why most of the loss is an accounting charge
When a startup raises money through instruments that can convert into shares, it often has to record them at fair value. As the company’s valuation climbs, so does the value of what investors could convert into, and the increase is booked as a loss. Anthropic’s valuation rose to $965 billion by May 2026, so the charge was large. It is a real cost to existing shareholders through dilution, but it is not cash out of the door.
The chart splits the net loss in the Anthropic prospectus using Reuters’ figures: roughly $34 billion of accounting charge and roughly $8 billion of everything else, alongside the operating loss for each year.
The operating loss is the number to watch
The operating loss is the cleaner measure of the business, and it grew 2.7 times, from $2.98 billion to $8.06 billion. The arithmetic is simple. Revenue of about $4.6 billion against operating expenses of $12.65 billion leaves a gap of about $8 billion, which matches the reported loss. The critic Ed Zitron put it bluntly on X: Anthropic “Spent $12.6bn to make $4.6bn in revenue in 2025”.
Growth: The Other Half of the Anthropic Prospectus
The same Anthropic prospectus that shows the losses also shows why investors are queuing. Revenue rose twelvefold in 2025 to nearly $4.6 billion, which implies roughly $390 million in 2024. Then growth accelerated. The FT reports that second-quarter 2026 revenue alone reached $11.5 billion, and that Anthropic is on track for its second straight quarter of operating profit on an adjusted basis.
Putting the quarter in context
One quarter of 2026 brought in about 2.5 times the revenue of the whole of 2025. Reuters reported earlier in September that second-quarter revenue a year earlier was $787 million, so the year-on-year increase for the quarter is roughly 14.6 times. The chart shows the scale, using the reported figures and the 2024 estimate derived from the twelvefold growth.
What drove it
Fortune points to products like Claude Code, which “spurred a sudden growth in users” and left Anthropic in “a compute crunch”. Enterprise demand for AI models that write code and run agents is the core of the growth story, and it is also why the next section, on costs, matters so much.
Compute Costs and $518 Billion of Commitments in the Anthropic Prospectus
The Anthropic prospectus puts 2025 spending on compute and infrastructure at $7.33 billion, three times the 2024 figure and, by our arithmetic, 58% of its $12.65 billion of operating expenses. Looking ahead, the Anthropic prospectus lists $518 billion of cloud, computing and infrastructure obligations “in coming years”. This year alone it has signed compute deals with Google, SpaceX and Nscale, and we reported on its $11.6 billion cloud deal with Akamai.
Cash on hand against commitments
The contrast is stark. Anthropic had $20.28 billion of cash, cash equivalents and short-term investments at the end of 2025. Its commitments are about 25.5 times that figure. The gap will be closed by revenue, by the IPO proceeds and, very probably, by debt, and investors will want to know the timing of each commitment before they can judge the risk.
What the sceptics say
Not everyone is convinced. The equity analyst Ross Hendricks asked on X: “Just a casual mention of raising half a trillion in a single year, really? Does anyone believe this? Let’s say they pull $100B in equity from the IPO… and then what, $400B from the debt markets?” The Anthropic prospectus apparently spreads the obligations over several years, but the scale of the question stands.
Two Customers and No Long-Term Contracts
The FT reported that the Anthropic prospectus shows nearly a quarter of 2025 revenue coming from just two customers, whom the reports do not name. On revenue of about $4.6 billion, that is roughly $1.15 billion between them. Reuters adds that, in its risk factors, Anthropic warned that “many of its largest clients were not locked into long-term contracts and could cut or stop spending.”
Why this matters more than it looks
Concentration is common in young enterprise software companies, but it is unusual at this scale and valuation. If either of the two customers moved a large share of spending to a rival model, or built more in-house, the effect on revenue would be immediate. That risk sits alongside the price war between frontier labs, which has pushed model prices down sharply this autumn.
The Warning That Its AI Could End Humanity
The part of the Anthropic prospectus that drew the most attention is the risk section. According to the reports, it warns that advanced AI could pose “catastrophic or existential risks to humanity”, and it describes behaviours that Anthropic says its models have shown or could show. The table lists the specific wording reported by the FT, Reuters and the Guardian.
| Reported wording | What it refers to |
|---|---|
| “catastrophic or existential risks to humanity” | The most severe outcome of advanced AI |
| “self-preserving behaviours” | Models acting to protect their own operation |
| “resist shutdown” | Attempts to avoid being turned off |
| “conceal or manipulate information” | Hiding or distorting what the model has done |
| behaviour “resembling blackmail” | Coercive behaviour seen in controlled tests |
| “significant limitation” | A model that knows it is being tested is harder to assess |
The company’s own words
The Guardian quoted one line directly: “Our development of highly advanced models, platforms, and applications and expansion of use cases could further increase the risk that our models cause harm.” Reuters noted that the warnings in the Anthropic prospectus reflect evidence from Anthropic’s own research that increasingly autonomous models “can behave in unexpected and potentially harmful ways, including sabotaging code, assisting fraud and manipulating information in controlled tests.”
A first for an SEC filing
TechCrunch said the phrase “existential risks to humanity” appears to be a first, “judging by a quick scan of the SEC’s database”. Companies routinely warn about cybersecurity, safety, regulation and competition. Warning, as the Anthropic prospectus does, that the product could contribute to human extinction is new.
Why Risk Factors Fill 80 Pages of the Anthropic Prospectus
Reuters counted roughly 80 pages of risk factors in the 261-page main body of the Anthropic prospectus, against 48 pages describing the business. The chart shows the split. The remaining 133 pages cover everything else, from the offering to management and the financial discussion.
What the law asks for
US securities rules require a prospectus to disclose the material risks of investing, and lawyers err on the side of saying too much, because an undisclosed risk that later happens invites lawsuits. We looked at how those rules interact with doom warnings in our earlier piece on AI doomsday warnings and IPOs. The Anthropic prospectus goes further than legal caution alone would need, which is consistent with what it says in public.
Pace the Frontier Meets the Public Markets
The Anthropic prospectus arrives in the middle of a public campaign by its chief executive. Dario Amodei has spent September calling for the industry to “pace the frontier”, arguing it “must slow the pace at which we improve the capabilities of AI models”, and, according to TechCrunch, told the UN Security Council that AI is “the most important global security issue facing the world today.” We covered the proposal in detail when Amodei set out his pacing plan.
The debate inside and outside Anthropic
The Guardian notes that the debate intensified after an Anthropic researcher, Jacob Coxon, resigned warning that people building AI “earnestly believe that it could kill us all by the end of the decade”, and a senior safety researcher publicly put the chance at more than 10%. Critics say such warnings are unverifiable. Dan Ives of Yorkville Ives told CNBC the slowdown calls are a “head scratcher”, arguing that “if they slowed down, China would just accelerate and win.”
Selling risk and safety at the same time
There is a coherent reading of all this. Anthropic has always sold itself as the safety-focused lab, and enterprise buyers have rewarded that. Telling investors in detail what could go wrong is consistent with that brand. The tension is that the same document asks investors to fund $518 billion of compute to build more capable models.
What the Anthropic Prospectus Implies About a $2 Trillion Valuation
Backers reportedly believe the company could list above $2 trillion, more than double its $965 billion valuation in May. Nasdaq Private Market puts its current private-market value at $1.32 trillion, according to Stocktwits. The table shows what $2 trillion means against the revenue figures in the Anthropic prospectus.
| Comparison | Arithmetic | Multiple |
|---|---|---|
| $2tn against 2025 revenue | 2,000 ÷ 4.6 | About 435x |
| $2tn against Q2 2026 revenue annualised | 2,000 ÷ (11.5 × 4) | About 43x |
| $2tn against the May valuation | 2,000 ÷ 965 | About 2.1x |
| $2tn against SpaceX at its June IPO | 2,000 ÷ 1,770 | About 1.1x |
Which multiple is fair
The 435x figure looks absurd but uses a year that is already out of date. The 43x figure on annualised second-quarter revenue is high, but not unheard of for a company growing this fast. The real question is whether the growth in the Anthropic prospectus continues, and that depends on the customer and compute risks above. Reuters notes that SpaceX, which listed at $1.77 trillion in June, sold shares at $135, rose 19% on its debut and now trades around $147.
Governance on top of valuation
Buyers of the shares will also accept limited control. We reported that Anthropic plans to give its co-founders 50.1% voting control, and that it has chosen a Nasdaq listing. Public investors will be funding a company whose founders keep the final say on how fast to build.
What the Anthropic Prospectus Means for Businesses Using Claude
For customers, the Anthropic prospectus is unusually useful. It is the most detailed view yet of a key AI supplier’s finances and its own view of the risks. Treat it as vendor due diligence.
Supplier resilience is good, with caveats
$20 billion of cash, fast-growing revenue and quarterly adjusted operating profit make Anthropic a durable supplier in the short term. The long-term risk is the $518 billion of commitments. If growth slows, a supplier under that much pressure may raise prices, change terms or prioritise its largest customers.
Plan for concentration on both sides
If two customers make up a quarter of Anthropic’s revenue, smaller customers have less influence over roadmaps and pricing. Keep your prompts, agents and data portable, test at least one alternative model, and avoid building critical processes that only one provider can run. A clear AI strategy should name that fallback.
Read the Anthropic prospectus risk factors as a safety checklist
The behaviours Anthropic lists, from concealing information to resisting shutdown, are the ones your own controls should assume can happen with any frontier model. Approval gates, logging and limited permissions for AI employees and autonomous agents matter more, not less, when the vendor says so in a legal filing.
What Happens Next to the Anthropic Prospectus
A company that submits confidentially must file its registration statement publicly at least 15 days before its roadshow begins. That public filing will contain the full Anthropic prospectus, with audited financial statements, and will let analysts check the figures reported so far. Reuters reports that the listing is likely to come after the US midterm elections in November, later than the October window we described when the Anthropic IPO slipped to mid-October.
What to look for in the public Anthropic prospectus
Four things: the schedule of the $518 billion of commitments by year, the identity or type of the two large customers, the reconciliation behind “adjusted” operating profit, and whether the existential-risk language survives legal review in the same form.
Anthropic Prospectus FAQs
Has Anthropic filed its prospectus publicly?
No. As of 29 September 2026, the document is a confidential draft that Reuters and the FT have reviewed. It will be made public at least 15 days before the IPO roadshow.
How much did Anthropic lose in 2025?
The Anthropic prospectus reportedly shows a net loss of about $42 billion, of which about $34 billion is a non-cash accounting charge on convertible financing. The operating loss was $8.06 billion.
How fast is Anthropic growing?
Revenue rose twelvefold in 2025 to nearly $4.6 billion, and second-quarter 2026 revenue alone was $11.5 billion, according to the FT.
What does the prospectus say about existential risk?
It reportedly warns that advanced AI could pose “catastrophic or existential risks to humanity” and lists behaviours such as attempts to “resist shutdown”, to “conceal or manipulate information” and behaviour “resembling blackmail”.
When will the Anthropic IPO happen?
Reuters reports it is likely to come after the US midterm elections in November 2026. No date has been announced.
References
Anthropic’s IPO prospectus shows sweeping AI vision, surging costs (Reuters)
Anthropic’s IPO prospectus shows sweeping AI vision, surging costs (CNBC)
Anthropic warns of existential risks to humanity in IPO prospectus (Financial Times)
Anthropic warns of existential AI risks to humanity in IPO document (The Guardian)
Anthropic’s leaked IPO prospectus details steep losses, rapid growth (Fortune)
Anthropic warns investors of AI’s existential risk to humanity (CNBC)
Anthropic’s IPO filing shows soaring revenue, mounting costs, and existential risks (The Decoder)
Anthropic IPO prospectus leaked: $42B loss (Stocktwits via Yahoo Finance)
Enhanced Accommodations for Issuers Submitting Draft Registration Statements (US SEC)
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
